Closed gijzelaerr closed 10 years ago
Original comment thread migrated from bugzilla
So far this only affects Debian lenny (8.10 packages), but I'd like to get to the root of this problem as it's VERY annoying.
Got it, the mystery is why it works right in the first place on most platforms. It all boils down to a float mtime:
purr(Purrer.py:120:isUpdated): WatchedFile /home/oms/Waterhole/contrib/OMS/QMC2/Jul21/batch_cal.sh is updated: mtime 1279968354.779170 07/24/2010 12:45:54 PM, old mtime 1279968354.000000 07/24/2010 12:45:54 PM
Because the timestamp gets converted to an int when written to a purrlog, if the filesystem reports a fractional timestamp, the file will always appear to be newer! Perhaps this is filesystem-dependent? xfs reports fractional timestamps.
I have implemented a fix on birch (basically, keep all timestamps as floats now), will test this and commit later.
Fixed in r 7836.
at 2010-07-13 15:59:09 Oleg Smirnov reported:
Already archived files picked up again upon restart of Purr